1. Field of the Invention
This invention pertains generally to digital representation of electrical signals, and more particularly to oscilloscopes and signal processing for removing distortions in signals digitized using a Time Stretch Analog to Digital Converter (TS-ADC).
2. Description of Related Art
The ability to capture and analyze waveforms is critical for numerous apparatus and systems. At the center of this capturing ability are techniques for converting analog signals to digital signals within Analog-to-Digital Converters (ADCs).
Over the past few decades continued advances in the state of digital technology have made it possible to perform impressive amounts of signal processing and computation within the digital domain. Digital information provides a number of benefits, such as convenient storage (e.g., electronic, magnetic or optical memories) and reproduction. These advantages have brought about what is often referred to as the ‘digital revolution’ to enable easy storage, processing and transmission of vast amounts of information. However, most signals in the physical world are analog in nature requiring analog-to-digital converters (ADCs) to translate these signals to a digital representation. ADCs, therefore, are crucial components for the majority of modern day electronic devices. It is natural, therefore, that the demand for lower power, faster and higher resolution ADCs is continuing to grow.
Ultra-wideband high-resolution ADCs, in particular, are required for a number of extremely important applications. For example, to satisfy the increasing bandwidth demands for Internet backbones and data centers, optical links are being tested which achieve data rates as high as 100-Gbps per wavelength channel. Examples include the 100G Ethernet for local area networks and the 100G transport for long-haul portion of the networks. These links are configured for using advanced modulation formats to achieve high spectral efficiencies. To demodulate these signals and to correct for the impairments suffered by data after it propagates through the optical transmitter, fiber and receiver, the receivers in these extremely fast optical links require ADCs with bandwidths in the tens of GHz range. High-speed ADCs are also increasingly finding applications in advanced test and measurement equipment, radar systems, software defined radios and biomedical instrumentation. Test and measurement equipment is required for capturing high-speed signals up to these 100-GHz frequencies, for example, the signals used in millimeter-wave radios and vehicular radar systems.
The majority of ADCs used today in digitizing electrical signals are based on CMOS electronics. Architectural improvements and availability of faster devices have made it possible to extensively rely on signal processing and calibration techniques, which has improved the performance of these ADCs. However, these improvements cannot fully satisfy the growing demand for wide-bandwidth high-resolution ADCs. Certain fundamental issues also exist associated with the use of the electronics, such as error in the sampling clock referred to as aperture jitter, and the finite switching time of transistors that lead to an error referred to as comparator ambiguity. These fundamental issues limit the speeds of electronic ADCs, or at a given speed limit their resolution. Therefore, it becomes imperative to use other technologies such as photonics, to overcome these inherent limitations of electronics. Photonic Time Stretch Analog-to-Digital Conversion (TS-ADC), such as represented by FIG. 3, is a technique previously developed by the Applicant (e.g., U.S. Pat. No. 6,288,659) that uses optics to slow down high speed analog signals in time by manipulating the signal time scales, allowing them to be captured by electronic ADCs operating at below the capture speeds.
FIG. 1 illustrates a block diagram of a photonic time-stretch pre-processor 10. The system is effectively a dispersive analog optical link having an optical source 12, Mach-Zehnder (MZ) modulator 24, disperson element 26 and photodetector (PD) 28. The continuous wave laser source of the link is replaced for this application with a chirped pulse source 12, such as having a Mode-Locked Laser (MLL) 14, Highly Non-Linear Fiber (HNLF) 16, supercontinuum pulse 18 providing a signal, shown at node A, which is passed through dispersion fiber 20 for outputting chirped pulse 22 at node B.
Prior to the photonic time stretching, the electrical input signal 23 is received by a modulator 24 which controls the modulation of the linearly chirped optical femtosecond (fs) mode-locked laser pulses as seen at node C. Propagation of these pulses through dispersive fiber 26 stretches the modulated pulses with respect to time as seen at node D.
A photodetector (PD) 28 converts these optical signals back to the electrical domain and the resultant electrical signal is a stretched replica of the original signal with significantly reduced analog bandwidth, as represented by signal 30 shown in the figure. This signal can now be recorded by a real-time electronic digitizer. It should be appreciated that only as a result of the time-stretching, does the ADC have sufficient bandwidth to convert the signal, as the ADC by itself is not fast enough to capture the original signal. Time stretch factors of up to 250 have been achieved with this approach, and electrical signals with frequencies as high as 95-GHz have been digitized in real-time at 100-fs intervals using the TS-ADC.
It should be readily recognized, however, that time stretch preprocessing can add distortion to the signal being recorded. The sources of this distortion include, but are not limited to, dispersion penalty that creates fading or nulls at certain signal frequencies, intermodulation and harmonic distortion caused by the fundamentally non-linear transfer function of the electro-optic modulator, dispersion induced non-linear distortion, non-linear distortion caused by saturation of the photodetector at high optical powers, distortion caused by nonlinear optical interaction within the dispersive medium, and non-uniform, time-dependent, stretch factors which create time-warps.
FIG. 2 depicts time stretched signals with: (1) the solid trace showing an ideal signal without time-warp; (2) dashed traces showing a signal with time-warp; (3) trace with squares represents error; and (4) the trace with circles representing amplitude of the time-warp.
Bandwidth limitations caused by dispersion can be overcome using several techniques, such as phase diversity, single sideband modulation, and equalization techniques. Prior work by the Applicant has shown that significant amounts of non-linear distortion due to various sources can be suppressed by differential operation and arcsine correction. However, non-linear distortion added by dispersion effects, Mach-Zehnder modulator (MZM) transfer function and MZM wavelength dependent bias offsets comprise the dominant portions of the residual non-linear distortion.
FIG. 3 illustrates a photonic time stretch ADC in which an RF electrical signal is segmented and stretched, digitized by slow backend digitizers, and then finally combined to obtain the digital copy of the original signal. The figure also depicts changes in the signal spectrum due to the stretching.
In the figure is shown a mechanism 50 for capturing the time-stretched signal continuously, by dividing the RF signal 52 into multiple segments 56a-56d (e.g., which may comprise any desired number of segments) using wavelength division demultiplexing (WDM) 54 (e.g., filter(s)). The use of WDM assures that the stretched signal segments do not overlap and intermix with each other after stretching. The time-stretched and slowed signal segments are then converted to digital samples by slow electronic ADCs 58a-58d. It should be appreciated that the term “slow”, as used in this respect, only means a device that has insufficient speed for directly converting RF signal 52. Finally, these samples are collected 60 by a digital signal processor (DSP) and rearranged 62 in order to get the output signal as the digital representation of the original analog signal. Any distortion added to the signal by the time-stretch pre-processor is also removed by programming within the DSP (e.g., any processor or combination of processors which process the digital signals) performing according to aspects of the present invention.
Devices for capturing signal waveforms, such as digitizers and oscilloscopes, rely on analog-to-digital conversion. At its current state of the art, digitizers and scopes (oscilloscopes) are categorized into two classes. Equivalent-time digitizers (or sampling oscilloscopes) rely on repetitive or a clock synchronous nature of the signals to reconstruct them in time. In a sampling oscilloscope, the signal is sampled at MHz frequencies (typically 100-kHz to 10-MHz) and then reconstructed digitally, requiring a long time to obtain the original signal with high fidelity. While they can reach equivalent time bandwidths of up to 100-GHz, they are not capable of capturing non-repetitive waveforms. Even for repetitive signals, they cannot provide real-time information about the dynamics that occur at rates faster than a few MHz. Equivalent-time sampling is similar to the strobe light technique used for measuring cyclical events which are much faster than the speed of the detector. For example, periodically flashing a strobe light on a vibrating tuning fork can make it appear to be vibrating very slowly and can be used for studying vibrations which are too fast for the human eye to discern.
The second type of digitizers, called real-time oscilloscopes, continually sample the signals as they change, but have input bandwidths limited to only a few GHz. The fastest practical real-time oscilloscope currently available has a bandwidth around 20 GHz.
Accordingly, a need exists for a system and method of converting fast waveforms into digital representations while not generating significant distortion. These needs and others are met within the present invention, which overcomes the deficiencies of previously developed ADC systems and methods.